British AI and autonomous driving tech company Wayve has claimed that its algorithm is learning to drive like a human being – using just cameras and a satnav, rather than an onboard sensor suite and hand-coded rules.
In a blog post on its website, the company claims that a different approach is needed to autonomous driving, especially on European roads, which tend to be narrower, and less straight and predictable than the highways in or between American cities.
The company claims its end-to-end machine-learning system outperforms rival autonomous driving systems, even on roads that test vehicles have never encountered before. In the blog post, it shared a video of a test vehicle learning from test-driver interventions on unfamiliar roads in Cambridge.
The company believes that electric, autonomous vehicles are the future of transport – a view echoed by many innovators in the tech and automotive sectors. However, Wayve says this is only possible using end-to-end deep learning, rather than what it sees as expensive, bulky, sensor-based solutions, and self-driving systems that need to be modified and adapted to different locations.
Wayve claims its system is available at roughly one-tenth of the cost of more ‘traditional’ driverless approaches.
However, some have questioned whether the cost of onboard hardware is the primary challenge facing an industry that is awash with investors’ cash. Indeed, the company’s claims have been greeted with scepticism and even outrage by rivals, who have pointed out the falling cost of LiDAR, radar, and other technologies.
“It’s lunacy,” Rick Tewell, COO at Velodyne LiDAR told website Mashable, which broke the story from Wayve’s blog. “AI performs a lot better with a lot more data, than less data.”
Leilei Shinohara, VP of R&D at RoboSense, added, “Sensors can get details that the human eye can’t.” Driverless cars may only use LiDAR five percent of the time in certain situations – if radar fails or is incomplete – but those occasions could be critical, he explained.
“The whole point of self-driving cars is to be be safer than a human driver,” Matt Weed, Luminar Technologies’ director of technology strategy, told Mashable. “Why you would eliminate technology that maximises safety inputs doesn’t compute. You want to be able to get as much good information about the world as you can.”
With one death for every 1,000 traditional cars on the road worldwide every year, making transport safer is a global challenge.
That Wayve’s technology is disruptive is not in doubt. But has the British company hit upon a simpler, cheaper, smarter solution than the likes of Tesla, Waymo, GM, Uber, and Apple? Or is it making a dangerous gamble with human beings’ safety on the road?
Let us know what you think.